CN101790726B - Monitoring methods and apparatus - Google Patents
Monitoring methods and apparatus Download PDFInfo
- Publication number
- CN101790726B CN101790726B CN2008801009171A CN200880100917A CN101790726B CN 101790726 B CN101790726 B CN 101790726B CN 2008801009171 A CN2008801009171 A CN 2008801009171A CN 200880100917 A CN200880100917 A CN 200880100917A CN 101790726 B CN101790726 B CN 101790726B
- Authority
- CN
- China
- Prior art keywords
- functional group
- functional
- sensor
- data stream
- sensed
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0208—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the configuration of the monitoring system
- G05B23/021—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the configuration of the monitoring system adopting a different treatment of each operating region or a different mode of the monitored system, e.g. transient modes; different operating configurations of monitored system
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
- G05B23/0254—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
Abstract
A method of monitoring an evolving system, the method including the steps of : obtaining a plurality of sensor data streams relating to outputs from sensors monitoring said system, wherein at least one of said sensors monitors a condition of said system, and wherein at least one of said sensors monitors a causal agent for said condition; iteratively constructing a plurality of functional nests, each functional nest being a functional formed from a combination of selected functionals from a basic set of functionals; determining an output data stream for each functional nest by inputting said sensor data streams into said functional nests; selecting a functional nest from said plurality of functional nests based on said output data streams; and using said selected functional nest to monitor said system.
Description
Technical field
The present invention relates to the method and apparatus for monitoring Evolution System state.
Background technology
Wish under many circumstances the state of monitoring Evolution System (evolving system), and diagnosis and prediction service about system are provided when system's Temporal Evolution.These situations for example comprise the structure of monitoring buildings for example or vehicle and/or the process of monitoring for example industry or environmental process and so on.The result can be for assessment of system performance, and can the standby maintenance system and get involved to improve performance, stability in early days, prevents fault or takes some other preventative or remedial actions.
Substantially there are two kinds of known ways to monitor Evolution System: the statistical study of differentiation or based on the process model of " priori (a-priori) ".
In the mode based on " priori ", the mechanism of research, understanding system bottom, for example physics and/or chemical process and corresponding modeling.How input initial parameters and moving model will begin Temporal Evolution from initial conditions with display systems in the model.
Be that based on a problem of the mode of " priori " the bottom process is not to be known always or to obtain fine sign, usually may be along with the time changes.
In the mode based on statistics, analyze the data from the time that depends on of system, to determine significant pattern on the statistical significance.This mode does not need to understand first floor system mechanism, but must want conscientious result, and the result usually only sets up under confined condition.
Current needs provide new mode to monitor Evolution System, and new mode can be assisted system is made diagnosis and/or prediction.
Summary of the invention
One aspect of the present invention provides a kind of method of monitoring Evolution System, and the method comprises the steps:
Obtain with from the relevant a plurality of sensor data streams of the output of the sensor of monitoring said system, the situation of at least one the described system of Sensor monitoring in the wherein said sensor, and the reason main body of the described situation of at least one Sensor monitoring in the wherein said sensor;
(iteratively) makes up a plurality of functional groups (functional nest) iteratively, and each functional group is by the functional that is combined to form of concentrating the functional of selecting from fundamental functional;
By described sensor data stream being input in the described functional group, determine the output stream of each functional group;
From described a plurality of functional groups, select the functional group based on described output stream; And
Utilize selected functional group to come monitoring said system.
Functional is defined as the function of function and the function on the vector space.The vector space of input is the data stream of one or more time encodings, and the vector space of output is the data stream of single time encoding.In this case, input traffic will be the time dependent value that is monitored by system sensor of the reason main body of the situation of monitored system or state, and output will be system status or state differentiation (evolution) in time.
Functional is including, but not limited to all solutions of ordinary differential equation.The functional of functional is functional.The functional group is from oneself any combination as the functional the baseset of functional.
A kind of method that comprises above-mentioned steps can select representative system best how various sensed parameters (reason main body) to be made in time the best functional combination (best functional group) of response.For example, system may relate to the corrosion of physical arrangement, and sensor data stream can be the value for one or more positions corrosion main body, such as wetness value, pH value etc., and output stream can be that structure has the tolerance how to approach apart from fault.
Use functional can make up based on " priori " and based on the best aspect of the modeling of statistics, because procedural knowledge will be embedded in the form of functional in the baseset and how make up in them, but select best functional combination (with they bottom processes with reflection) to be driven by sensor, the relation of the functional combination representative of therefore selecting is the direct response to physical presence situation in the system.
The method can be regarded as relevant with system evolution potentially theory is divided into one group of (being represented by functional) module, and utilize the data of sensed system that these modules are reassembled into workable system model (selected functional group).Then this model can be used for diagnosis and/or the prediction of system.
The functional that makes up can be taked much multi-form, can change to some extent between the system and between the monitored situation from its fundamental functional collection that forms functional combination, thus reflect may estimate in system, to occur and the bottom relevant with monitored situation machine-processed.Can be split into by the theory that will be referred to system action single functional and construct the fundamental functional collection, the method can be reassembled into these functionals the new theory of whole system.
The selection of functional group will roughly export to drive by optimizing it for monitored situation.Reason body data stream is input in the various possibility functional groups, selects the data stream of output stream and monitored situation to have the functional group of best-fit.Can carry out multi-dimensional optimization by the coefficient to them and select the functional group.
Functional can be linked to specific data stream, particular sensor for example, but preferably, data stream can be applied to any functional.In this case, will be from the composite construction functional group from selected sensor data stream of being combined of selected functional.
Functional can not have, has one or surpass a coefficient, and to one or more data flow operations.It may be useful that functional in the baseset is restricted to that those have a coefficient and a data stream or do not have coefficient and the functional of two data stream is arranged, because this can realize the simplicity of programming.Also preferred at least one functional acts on and surpasses a data stream, because the result's (for example temperature in the corrosion monitoring system and moisture) from standalone sensor can be coupled like this.
Preferably, the functional in the baseset has dissimilar non-linear.This can realize good numerical property, and can avoid repetition and cover on a large scale possibility theory.Typical non-linearly can comprise exponentiation, integration, threshold process, multiplication sensor data stream, interpolation, quantization, application preassignment function and utilize the differential equation to delay time or phase shift.Can square come the mandatory adoption positive coefficient by what get them when needed, for example, cause numerical problem and cause numerical algorithm constringent unstable because of the number divided by zero passage in order to prevent.
More different each other between the functional, computing machine just can more quickly be distinguished them.
Can in processor storage, store simultaneously, for example keep a plurality of different functional groups, and the sensing data in can utilizing during the special time before selecting to be used for the functional group of monitoring is optimized the coefficient of each functional group.
Can be along with the time be upgraded selected functional group, the functional group that for example keeps in the processor storage, for example, after receiving each new value of data stream or after receiving the new data of set amount.Although may consider all possible functional group when each the renewal, this needs sizable computing power.In order to reduce this requirement, can might determine first group of best-fit functional group the functional group from institute, then from this group, select best functional group.When new data enters, can come this group best-fit functional group of frequent updating by optimize functional group coefficient for new data, and can constantly select new best functional group from this best-fit group.Can by might the functional group in again search upgrade more erratically the set of the best-fit of the functional group that keeps in the processor storage.
Can functional be combined into the functional group by any appropriate ways.Can use the discrete optimizing method that is suitable for solving NP difficulty (uncertain polynomial time difficulty) problem, in order to functional group quantity is remained on manageable level, also providing simultaneously can be to the functional group of monitored system Accurate Model.Can use branch-bound method.For example, the method can for example be utilized the combination of all possible functional and all possible functional of all possible data stream combination formation pair, and can utilize the multidimensional unconstrained optimization to come to optimize these functionals pair for the functional coefficient.Then can be in might making up, for example on each nested level, to combination, then can be that their coefficient is optimized full income functional group with each fundamental functional and each functional.Then can be for best-fit to the functional of such formation to sorting with the functional tlv triple, by abandoning the poorest functional group of match the quantity of functional group is dropped to maximum quantity.Then can be by repeating this process to other functionals of increase such as each tlv triple, until reached the nested level of expectation or until satisfied some other standards, for example height match.Then can with the set of the functional group that the obtains basis with the best functional group that elects, perhaps can use the functional group of the best-fit of limited quantity in them.
In order further to improve the processing time, can provide rule for the combination functional, so that various unsuitable combination can not occur.For example, if the combination of functional is of equal value each other, then only can allow one of combination.And, can avoid being difficult to the functional combination that numerical value is processed, because this combination may not relate to actual evolution mechanism.
Another kind of standard can be if the output stream that the functional group provides does not meet the expection feature of monitored system, then to refuse the functional group.For example, maybe can not descend if monitored system status can not rise, then can be for example optimize the functional group or do the functional group best-fit check before refusal produce the functional group of the output of rising respectively or descending.This can for example prevent from using the functional group that alleviates for corrosion, because corrosion will only can rise or keep constant, can spontaneously not alleviate.
The best functional group that this method can be provided is used for diagnosis and prediction.In one form, in selected best functional group, input the prediction data stream in future that is used for the reason main body, in order to the output stream corresponding to the prediction future value of the relevant system status of functional group is provided.Prediction future data stream can be based on the data stream of previous sensing, and can for example relate in the data stream of sensing formerly and rerun all data values.So the reason main body of system status or state will usually be consistent, and can for example be constant or in a known way circulation, therefore can repeat simply old traffic value to determine new data value.Can also only use past data stream one section, for example, relate to a section of system's material circumstance, material circumstance for example is to be exposed to specific environment within the time of this section.Also can revise past data with for example expection by environment change to look ahead expection difference in the data stream.
Because the selected best functional group of system has reflected its bottom physical responses and mechanism, the selected functional group of functional group and other system (similar system that for example has diverse location, geometric properties, material or environment) can be compared, how similar to have between definite system.Comparative example is such as the similarity of the quantity that can relate to public functional in the functional group and their orderings.The time history that can also relate to the functional group, for example how they are along with the time changes.So, can store in time functional group data, can compare the functional group data of two systems to determine the correspondence between the system.For example, system can relate to the position on the corrosion structure.
Similarly system can inform each other that they are to the selection of functional group.For example, if two systems are similarly, one is in than another and makes progress faster state (more advanced state), can use the functional group of selecting for the system that makes fast progress so to the slower system modelling of progress the time.For example, can utilize the current best functional group of the slower system of current best functional group interpolation progress of the system of making fast progress, make progress the functional group of development of slower system in stage of the better very fast system of predicted evolution in order to produce.Attempting to predict that not this interpolation may be particularly useful in system in the future but a little later when the fault of time or some other states recently (at this moment, the best-fit functional group of system may be than some variation of functional group of current selection).
During the functional combination in determining the slower system of progress regular, the history of the system that makes fast progress also may be useful, and is because can identify the trend of best-fit or the poorest match functional group, then can be in the slower system of progress preferred or avoid.
Also the result of monitored system can be applied to not monitored system.For example, the monitored system such as physical arrangement can comprise the subsystem that a plurality of quilts are monitored separately, for example structural position.In this system, owing to the reason of cost or practical problems (for example the inaccessible position of sensor, for example the slit of structure is medium), possibly can't monitor all subsystems.
In this case, can determine best functional group for sensed subsystem, can be used to best functional group and the not sensed subsystem of sensor data stream monitoring from sensed subsystem.So, can identify the reason main body is had similar response (for example in structure, for example geometric properties and material aspect are similar) sensed position and not sensed position, thereby they can use identical best functional group, and identical sensor data stream can be used in the sensed and not sensed position (for example position adjacent one another are) that is identified as having the reason main body of similar value.
Can revise functional group and the data stream of sensed subsystem (for example position) by predetermined extrapolation functional, in order to consider the expection difference of functional group between the sensed and not sensed position or data stream.Also can carry out interpolation to functional group or data stream from the sensed system of two or more similar type, think not sensed system, for example be positioned in the middle of two similar sensed positions or near not sensed position, functional group or data stream are provided.
For example, for the purpose of interpolation, the similarity of two systems may require the similarity of best functional group and/or their known intrinsic property (for example geometric properties and material composition).Can pass through the environment similarity, the similarity of two systems aspect reason body data stream is determined in structural adjacent position for example.
The present invention also is extended for the equipment of said method.Another aspect of the present invention provides a kind of equipment for the monitoring Evolution System, comprising:
The one group of sensor that is used for the parameter of monitoring said system, the situation of one of described sensor monitoring said system, and the reason main body of one of the described sensor described situation of monitoring; And
Processing module is used for:
From described one group of sensor receiving data stream;
(interactively) makes up a plurality of functional groups iteratively, and each functional group is the functional that is combined to form by the functional of selecting from the fundamental functional collection;
Apply described functional group to described data stream, to produce the output stream of each functional group;
Select the functional group based on described output stream; And
Utilize selected functional group monitoring said system.
One of sensor can monitoring system situation or state, at least one other sensor, but multisensor more normally will monitoring change the reason main body of monitored system state.
For example, for corrosion monitoring system, the reason sensor can comprise wetness sensor, temperature sensor, RH sensor, salinity sensor and pH sensor, and state sensor can be monitored corrosion current and can comprise linear polarization electric resistance sensor, galvanochemistry electric resistance sensor, galvanic couple and corrosion product sensor.Can also be with some state sensors as the reason sensor.
Sensor can be provided in together can monitor a plurality of positions in the sensor bunch of specific location in the system, each position has the cluster sensor.The functional group data of bunch position can be compared to each other, in order to determine whether the behavior of these position systems is similar, and such as the inherent feature of those position systems, for example the geometric properties of structure and material are such.
Can be used to the position that best functional group and sensor data stream monitoring from sensed position do not provide sensor.Data stream can be from the sensed position that approaches the not sensed position that for example has identical or like environment, and the functional group can be near position or at a distance, and the system that can depend on more is in the similarity of the inherent feature (for example their geometric properties and material) of these positions.
This equipment can be taked a lot of forms from the multi-purpose computer to the specialized processing units.This equipment can embed in the CPU (central processing unit), and CPU (central processing unit) is from a lot of different sensed system or subsystems, locations of structures for example, receiving sensor reading.It also can embedding distribution formula system in, bunch provide intelligent processing unit at each sensor, provide prediction etc. for the treatment of sensing data and for this position.These processing units can receive information from other processing units and from central controller, and central controller also can receive information from processing unit, in order to provide the overall situation to see and make up and monitor any global issue from the prediction of different processing units.
Except being provided by the real sensor system, the data stream that is used for functional also can or alternatively be formed by generated data.Can produce generated data with the output of Reality simulation sensor, generated data can come from or based on the actual sensor data.For example can use it for: the dry run on the equipment of not yet installing or assembling; In the space or interpolation or extrapolation from a processing unit to another processing unit; Extrapolation in time; Sub sampling for example produces data from daily measurement result with interval per hour; And replacement data that lose or that damage.
The method and equipment can be used for a lot of different application.For example, they can be used for monitoring structure and/or the process of monitoring such as industry or environmental process of buildings for example or vehicle.The result can be for assessment of system performance, and can the standby maintenance system and get involved to improve performance, stability in early days, prevents fault or takes some other preventative or remedial actions.
In an application that is particularly useful, the method and equipment can be used for the means of transport health monitoring, for example be used for aircraft industry.
The present invention also is extended for the software of carrying out the method for the present invention.The software that another aspect of the present invention provides a kind of equipment with being used for the monitoring Evolution System to use, described system comprises one group of sensor for the parameter of monitoring said system, the situation of one of described sensor monitoring said system, and the reason main body of the described situation of one of described sensor monitoring; And processing module, described software comprises a series of instructions, is used for so that described processing module:
From described one group of sensor receiving data stream;
Make up iteratively a plurality of functional groups, each functional group is the function that is combined to form by the functional of selecting from the fundamental functional collection;
Apply described functional group to described data stream, to produce the output stream of each functional group;
Select the functional group based on described output stream; And
Utilize selected functional group monitoring said system.
Any that should be pointed out that above-mentioned aspect can comprise any feature of above-mentioned any other aspect, and can comprise any feature of following any embodiment.
Description of drawings
Now with only also embodiments of the present invention will be described by referring to the drawings by way of example.Obviously, the specificity of accompanying drawing does not substitute the generality of above stated specification of the present invention.
In the accompanying drawings:
Fig. 1 is the schematic diagram that utilizes the method and apparatus of one group of functional monitoring Evolution System;
Fig. 2 is the process flow diagram of process that is identified for the best functional group of monitoring system;
Fig. 3 is the process flow diagram of determining therefrom to select for the process of one group of functional group of the best functional group of system monitoring;
Fig. 4 is the process flow diagram that utilizes the process of selected functional group prognoses system situation;
Fig. 5 can utilize Functional Approach at the schematic diagram of the structure of a plurality of sensed and not sensed position monitorings; And
Fig. 6 to 8 is the schematic diagram that can utilize Functional Approach each other structure of sensed and not sensed position monitoring at each.
Embodiment
With reference to figure 1, the system 1 of Temporal Evolution, for example the physical arrangement of regression usually is compound in itself.For example, in the system a plurality of machineries, physics and/or chemical process may occur, may be difficult to accurately know how in time how these processes carry out or each other mutual or their development.Therefore, may be difficult to utilize conventional modeling technique to this system modelling.
In the method, determine mathematic(al) representation or " theory " 2, various process/mechanism that it can express possibility and work in system 1.Then these theories 2 are divided into each fundamental functional g
1To g
n, each functional can be relevant with one or more processes or theoretical 2.
Functional can be considered as the function of function, it can act on the time code data stream of one or more inputs, to produce the output time encoded data stream.
These individual fundamental functional g
1To g
nBe stored in the storer 3 of processor 4.Processor 4 comprises data processing unit 4.1, primary memory 4.2 and secondary store 4.3.The software of a series of instruction types of primary memory 4.2 storages, to make the function of data processing unit 4.1 carry out desired, for example Fig. 2 is to function shown in Figure 4.That secondary store 4.3 temporary transient storage system 1 duration of works produce, the required data of processor 4 carry out desired functions.Processor 4 is with functional g
1To g
nBe combined into various possible arrangements (permutation) FN
1To FN
mEach of these arrangements is fundamental functional g
1To g
nIn one group of selected functional, can be considered as expression and how can work about system 1, be i.e. the possible mathematics " theory " of reason main body (causal agent) and the result status Relations Among that will monitor.
The data stream s that processor 4 uses from the sensor 5 in the system 1
1To s
1Select functional group FN
1To FN
mOne of as best functional group FN
o, how best functional group best prognoses system 1 makes response to the impact of sensor 5 monitorings, and utilize this best functional group FN
oDiagnose and/or prognoses system 1 Temporal Evolution how.
In order to determine best functional group FN
o, processor 4 can receive and not only relate to the reason main body but also relate to the as a result data stream s of system status
1To s
1For example, processor 4 can receive about the sensing data as the reason main body such as wetness, pH value, humidity, and can receive the sensing data of the as a result of system status that relates to the material corrosion state.Processor 4 then can be to functional group FN
1To FN
mVarious may the arrangement in input a variety of causes body data stream, and can determine which functional group provides the output best with the match of actual corrosion data stream.Then this functional group can be elected to be best functional group FN
o, and can be used for the monitoring and prediction System Development.
Along with receiving more data from system sensor 5, which functional group processor 4 can upgrade is best determining, therefore in response to the variation in the system 1, and best functional group FN
oCan temporal evolution.
Except arranging functional g
1To g
nTo provide outside the functional group, processor 4 can also be arranged sensor stream s
1To s
1So, although can be in it be used functional g
1To g
nBe limited to particular sensor so that select functional also must will use specific data stream, but preferably, functional can free action in any data stream that senses.So functional group FN will not only relate to the concrete combination of functional, and relate to the concrete combination of data stream and those functionals.
In addition, except the selection of optimizing functional and sensor data stream, processor 4 can also be optimized the coefficient of functional.For example, can with respect to the coefficient of making sensor output that the situation of just monitoring before the best functional group selection is associated and optimize each functional group, when selecting best functional group, can continue to optimize its coefficient according to new sensing data value.
Because the functional g that combines
1To g
nThe bottom physics or the chemical mechanism that roughly are fit to system 1 have been reflected, final theory (the best functional group FN that the method is selected
o) will comprise physics and/or chemical property, and since the best of breed of functional by the real sensor data-driven, so selected final functional group is the direct response to the actual conditions that exist in the system 1.Final theoretical FN
oTherefore will represent the theory that the sensor of how to work about system 1 drives, and with ratio as only more effective on physics and chemistry by the model of statistic processes driving.The actual mechanism of also having an effect in the understanding system in detail and obtain best functional group, and best functional group can be rapider than the response of " priori " model.
Fig. 2 is that processor 4 is in order to determine best functional group FN
oAnd the process flow diagram of a kind of exemplary method of carrying out.
In step S1, processor 4 is determined operable sensor data stream s
1To s
1With functional g
1To g
nBaseset.What reason main body and situation these will monitor based on expection and needs that what mechanism in the system 1 is had an effect.Can determine functional by the independent functional of various theoretical analysis that how to develop from monitored system, processor 4 reassembles into these " theories " newly " theory " that is combined to form from various functionals effectively.
The selection of functional can be limited to the functional of accepting a coefficient and a data stream or not accept coefficient and accept the functional of two data stream, thereby realizes being used for implementing the small routine of the method.For the numerical property (numerical performance) that obtains, what the functional in the baseset can specify different types is non-linear.Typical non-linearly can comprise exponentiation, integration, threshold process, multiplication (multiplying) sensor data stream, interpolation, quantization, application preassignment function and utilize the differential equation to delay time or phase shift.Can square come the mandatory adoption positive coefficient by what get them when needed, thereby reduce the risk of numerical problem.Functional is preferably to the data stream effect above, thereby coupling is from the result of different sensors.
In step S2, processor 4 is combined into a plurality of functional groups with functional and reason body data stream, and optimize the coefficient of functional in every group, so that realize the best-fit (best-fit) of output stream of the state sensor of the system status that will monitor with sensing from the output stream of functional group.
In step S3, processor 4 will realize that functional group selection with state sensor data stream best-fit is as best functional group FN
o, then store this functional group FN
o, to use during situation in the future in prognoses system.
The method of best-fit can be least square method, can comprise weighting, for example when the time series of the prediction of match system and actual monitoring value, can give closer value less weight for more outmoded value.This allows nearer error amount older value when prediction performance in future to play greater role.And when weighted value fully between the old times time was zero, this can help to improve computing velocity and reduce memory requirement.This also can realize tracker state (for example etch state) better, with the sudden change of answering system state (for example corrosion rate).Coefficient optimization can become can be by the problem of standard method solution in the multidimensional unconstrained optimization.
Along with new data value enters from sensor 5, can continue to determine best functional group FN with upper type
oYet, in order to reduce computing time, in step S4, processor 4 is determined best-fit N functional group finding in step S2, for example best-fit 100, and they are stored as foundation, are used for determining in the future best functional group by circulation execution in step S5 to S7.
So in step S5, the monitoring sensor data stream is optimized for the sensing data that upgrades from the coefficient of the highest N the functional group of step S4.Can carry out this renewal for every new data that receives or after receiving the new data value of setting quantity.After every suboptimization, select the best-fit functional group of N functional group as best functional group FN at step S6
oConstantly update best functional group FN when developing in system 1 in this way,
o
When system 1 develops in time, the initially the highest N functional group of step S4 may no longer be best-fit functional group, therefore at step S7, processor 4 can the Monitoring and Update situation, upgrade if exist, make process turn back to step S1, be combined with all possible functional of reappraising and redefine the highest N functional group.If there is no upgrade situation, this process continues to cycle through step S5 and S6.
The renewal situation can be any suitable triggering, can represent to reappraise to the fundamental group (basic group) of functional group.For example, can be after the time quantum of setting or received appearance after the new data value of set amount.Can also they be sorted by the best-fit according to the basic group of N, and when this ordering marked change, for example when a plurality of best-fit groups significantly change their positions in ordering, upgrade, thereby occur.
Therefore, the process shown in Fig. 2 has reduced the calculation requirement of processor 4, but still accurate best functional group is provided.
Fig. 3 is the process flow diagram of for example determining a kind of mode of best-fit functional group for the step S2 of Fig. 2 to S4.It has used the branch and bound search technology, in order to reduce computing cost.
In step S10, from functional g
1To g
nMight make up and reason data stream s
1To s
1Might be combined to form functional pair, and these functionals are optimized their coefficient in step S11.In step S12, in all suitable nested position with fundamental functional g
1To g
nAnother embed each functional in the group, wherein can control appropriateness by the rule that for example restriction repeats (duplication) and unsuitable system and/or numerical value behavior.The functional group of gained coefficient to them in step S13 is optimized.Then in step S14 according to they with sort to the optimization functional group of new formation with according to its group (in this first example, optimize comprise all functionals to) that forms them from the fitting degree of the data stream of state sensor, and in step S15, by abandoning the poorest group of match the functional group is limited in maximum quantity (for example 100).Can and abandon ordering and be combined into single step.The step S14 of Fig. 3 and S15 can be corresponding to step S3 and the S4 of Fig. 2.
Repeating step S12 is to the process of S15, until realize stop condition at step S16.So this process continues to add extra functional to each existing group, with the degree of depth of increase group, the quantity of restricted group is saved the processing time thus simultaneously.After suitable number of iterations, in step S16, stop to increase functional, the functional group set that existed at that time in step S17 storage is for example for the step S4 of Fig. 2.
This set of functional group can comprise the group that the degree of depth (namely wherein the quantity of functional) degree changes, as among the step S14, the functional group that is sorted not only comprises the group that has increased functional on it, also comprises those functionals than low degree that survived in the before this cycle.
The stop condition of step S16 can take on any appropriate form, and can for example embed the possible fundamental functional g of institute
1To g
nThe time, or occurred to expectation during depth capacity or decision set occurs when having than the more approaching best-fit of particular value nested.
Can to how making up fundamental functional g
1-g
nWith the limit calculation demand restriction is set.For example, can forbid only copying the functional combination of other functional combinations, the numerical value mode is processed or can not relate to physically significant mechanism because combination may extremely be difficult to.
And, can contrast the relatively output of functional group of the expection of monitored system status or known trend, and if its output do not meet expection trend and can refuse output.For example, may need the functional group that the output of monotone increasing or decline is provided.For example, when monitored situation can not rise or descend, for example can not descend and will only keep identical or when rising, may be suitable like this in corrosion and damage.
As mentioned above, in case determined best functional group FN
o, just can use it for the development in future of the monitored system status of prediction.This can be for example by the process implementation shown in Fig. 4.
In Fig. 4, at first at step S20, for example determine the best functional group FN of system by the process of Fig. 2 and Fig. 3
oNext, determine one group of reason body data stream in the future at step S21.This future, data for example can be the repetitions of previous sensing data, because reason main body value will usually have consistance to a certain degree, and can for example keep in time constant maybe can to circulate by the mode of understanding.Can also be the specific part of past data flow valuve with the Data Identification in future, the particular case that has for example experienced corresponding to system and again experienced in the future.For example, if at the corrosion condition of monitoring space vehicle, data can relate to given travel route etc.And data can be the previous sensor data streams of revising in the expection mode in the future, and for example, with known variation in the reflection system environments, it is medium that for example monitored structure moves to Different climate.
In case determined suitable data stream, just in step S22, they be input to best functional group FN
oIn, and at step S23 analysis result, for example to check important situation or indication such as fault when a certain amount of damage have occured and when may need remedial action.So then, can realize suitably arranging safeguarding etc., and can implement the prevention action to prevent fault etc.
Because best functional group FN
oReflect the bottom process that occurs in the system, therefore also can use best functional group FN by a lot of other modes
oFor example, can between the best functional group data of a system and another system (for example, the time history of selected best functional group, current best functional group etc.), make and intersecting relatively.For example, if the functional group of two systems is fully similar, can think that so system moves in a similar manner.Can use this information with a lot of modes.
For example, if a system is faster than another system evolves, for example degenerate faster, can notify the best functional group development of the slower system of progress with making progress the time history of the best functional group of system faster so.This may be particularly useful for further prediction in the future.Like this, although the best functional groups of determining according to Fig. 2 and 3 can be the good predict devices in future of arriving at once, it may not can keep relevant in future far away, for example, and when attempting to determine structure failure time etc.Therefore, by the identification similar system, best functional group data can be used for making progress faster system, with best functional group in the future of the slower system of better predicted evolution.
For example, can utilize the interpolation functional to be the make fast progress one or more best functional group of similar system of the best functional group interpolation of the slower system of progress.Then can use the best functional group of gained and estimate that reason body data in the future flows, with the development of the slower system of predicted evolution.
Can realize with discovery feature trend which position progress is determined faster by the state sensor that compares two positions and/or the best functional group history of observing them.And, whether similar in order to determine two positions, can compare their intrinsic features, for example geometric properties and material, rather than relatively or also can compare their best functional group data.
Also can the history of the best functional group of using the similar but system that makes fast progress in the rule of functional can be how made up in definition, thereby optimizing process can be accelerated.For example, can identify the trend in the system of making fast progress, when determining the best functional group of similar system, can preferably or avoid making up with the functional of those trend couplings.For example, can find that specific adjacent functional usually betides in the good or bad functional group of match, can preferably or avoid these functionals respectively.
Whether the best functional group data that can judge in several ways two systems are similar.For example, in the functional group, set the fundamental functional of quantity identical and when occuring with same order, and when each of these fundamental functionals acts on data from identical or similar sensor, can think that system is similar.
Also monitored system can be used for providing about the not prediction of monitored system.For example, as shown in Figure 5, system 1 can relate to the corrosion condition of structural detail 7, and monitoring system can relate to monitoring a plurality of subsystems, for example corrosion condition of ad-hoc location 7.1-7.5 in the structural detail 7.
This can be by providing reason main body sensor and state sensor and utilizing said process to realize to each position at all 7.1-7.5 places, position.
But, possibly can't provide sensor in all positions, for example, and only can be in some positions, sensor bunch for example is provided at 7.1,7.3 and 7.5 places.In this case, can use said process so that diagnosis and the predicted data to each position to be provided to each of these positions, can utilize best functional group and monitor not sensing position 7.2 and 7.4 from sensed position 7.1,7.3 and 7.5 sensing data.
For example, possible known location 7.2 and 7.5 pairs of reason main bodys will have similar response, for example because they are to be manufactured from the same material and to have an identical geometric properties, simultaneously may known location 7.2 be exposed to the reason primary influences identical with position 7.1, for example, because position 7.1 and 7.2 is in the A of first environment zone, and position 7.3,7.4 and 7.5 is arranged in second environment zone B.In this case, can be based upon the best functional group determined position 7.5 and the 7.1 reason body data streams that obtain from the position to the prediction of not sensed position 7.2.
Replace to use is can revise any one to consider the expection difference of best functional group between the sensed and not sensed position or data stream from the identical best functional group of sensed position or same data stream.This can be by realizing with predetermined extrapolation function, and this extrapolation function for example can be determined from laboratory experiment.This extrapolation functional can be inserted in the precalculated position in the previous best functional group of determining.
Also can use from the data stream and the best functional group that surpass a sensed position.For example, if estimating position 7.3,7.4 and 7.5 pairs of reason main bodys provide similar system responses, can use the best functional group that the interpolation functional obtains not sensed position 7.4 by the best functional group to sensed position 7.3 and 7.5 so.If need the difference between the compensated position, then the gained functional can also relate to the extrapolation functional.
When considering the appropriateness of interpolation, can make comparison to 7.3 and 7.5 best functional group whether similar to determine them.If think that they are similar, for example, if they have the identical fundamental functional that acts on the setting quantity that is in identical or similar order on the same or similar type sensor, can think that so sensed position has similar response, can think that also near the not sensed position two positions has identical response characteristic.In this case, if position 7.3 and 7.5 has similar best functional group, can think that they and position 7.4 all have similar response to the reason main body, can be the 7.4 best functional groups of distributing as the best functional group interpolation of position 7.3 and 7.5.
The data stream of position 7.4 also can be the interpolation from the data stream of position 7.3 and 7.5, because they are among the similar environment B.
Also these considerations and interpolation can be expanded to the three or more sensed positions of use.
Fig. 6 to 8 shows can be by sensed position monitoring other situations of sensed position not.Fig. 6 is the view of one jiao on structure, and Fig. 7 is the cross section by the slit, and Fig. 8 is the cross section by securing member.In Fig. 6, position 8.3 is jiaos of right-angle structure 8, may be not easy in this place sensors, and position 8.1,8.2 and 8.4 can be sensed or not sensed.Because position 8.1 and 8.2 is all on the vertical section 8a of structure 8, their environment can be similar, unless for example water drips above position 8.2, and the environment at 8.4 places may be different or be not difference, because it is on horizontal segment 8b, and because it is in the bight, the environment at 8.3 places will generally be different.
In this case, if only at 8.1 places sensor is arranged, 8.1 data stream can be directly used in position 8.2 and 8.4 so, can proofread and correct by predetermined angle (extrapolation) functional simultaneously and revise these data stream, in order to be 8.3 specified datas streams.Can be in advance definite angle corrected functional such as chamber experiment by experiment.
If at 8.1 and 8.4 places sensor is arranged, 8.1 data stream can be used for the data stream of position 8.2 so, and for position 8.3, can use the interpolation functional to 8.1 and 8.4 data stream, then can use the angle corrected functional to the interpolation result.
Fig. 7 shows the structure 9 that is formed by two part 9a that form slit 9c and 9b.In this example, the position 9.1 and 9.2 of outside, slit is arranged, can install there or sensor installation not, and the position within the slit 9.3, be difficult to there sensor installation.
In this example, if monitoring location 9.1 only, so can be directly from the position 9.1 data stream obtain the data stream of position 9.2, and can be by 9.1 data stream applies the data stream that slit extrapolation functional obtains position 9.3 to the position.This extrapolation functional will be again for example definite in the laboratory in advance, and its angle extrapolation functional from Fig. 6 is different.
If position 9.1 and 9.2 is all monitored, so can be by 9.1 and 9.2 data stream applies the interpolation functional and slit extrapolation functional obtains 9.3 data stream to the position.
Fig. 8 show securing member 11 and the metal decking 12 that engages between another kind of gap structure 10.Usually can not in gap position 10.1, place sensor, but the association of the possibility of the corrosion condition around the securing member 11 is very large, therefore in order to monitor the corrosion condition here, corrosion condition that can monitoring location 10.2, can use the data stream from 10.2, by apply another slit extrapolation functional (for example, in the laboratory, determining) to it, determine the corrosion condition at 10.1 places.
In Fig. 6 to 8, the sense position that can have from expectation meeting and not sensed position similar response is determined the not best functional group of sensed position.These sensed positions can approach not sensed position, for example can be the one or more of position shown in the figure, maybe can be structural other places.Equally, best functional group can be revised by extrapolation functional and/or interpolation functional, and will be applied to the predicted data stream that established data above flows to provide expection Corrosion developing in each position for example.
Generally speaking, method and apparatus of the present invention can be by providing the good fit functional combination to the direct monitored state of system, and the complication system of Temporal Evolution is carried out modeling.Within himself, embed true process and mechanism from the functional of its Selection and Constitute, and should combination be driven by True Data.Therefore, the functional combination that obtains will have bottom physics and/or chemical implication, and will make response to studied concrete system.
The functional combination of deriving not only can develop for passing through that they are predicted future for the reason body data stream in future of prediction, but also can provide the intersection between the system to compare, for example to determine how similar system has each other.
The method and equipment can be used for a lot of different application.For example, they can be used for monitoring the structure of buildings for example or vehicle and/or monitor the process of for example industry or environmental process and so on.The result can be for assessment of system performance, and can the standby maintenance system and get involved to improve performance, stability in early days, prevents fault or takes some other preventative or remedial actions.For example they can be used for the means of transport health monitoring, for example space vehicle monitoring.
Any suitable feature that they can be used for monitoring system, is transmitted the distance that product and means of transport travel from reaction vessel at for example cumulative degeneration (degradation) of structure.
This equipment can be taked much multi-form, can comprise any suitable sensor and data processing equipment, comprises multi-purpose computer or the dedicated hardware units of suitable programming.
This equipment can embed in the CPU (central processing unit), and CPU (central processing unit) is from a lot of different sensed system or subsystems, locations of structures for example, receiving sensor reading.It also can embedding distribution formula system in, bunch provide intelligent processing unit at each sensor, provide prediction etc. for the treatment of sensing data and for this position.These processing units can receive information from other processing units and from central computer, and central computer also can receive information from processing unit, in order to provide the overall situation to see and make up and monitor any global issue from the prediction of different processing units.
Obviously, can make various changes, increase and/or revise and not depart from the scope of the present invention previously described part, and in view of above instruction, can implement the present invention with software, firmware and/or hardware by the variety of way that those skilled in the art understand.
With regard to one or more applications in future, can be with the application as basis for priority, the claim of any this application in future can relate to any feature or the Feature Combination of describing among the application.Any this application in future can comprise one or more claims, and claims by way of example mode provide, and can advocate that aspect what be not restrictive in any future in the application.
Example
Example 1
This is based on the simple example that the manual car operating range is estimated, supposes following sensing data:
S
1Wheel rpm
S
2Engine rpm
S
3Gear (namely 1 or 2 or 3 etc.)
S
4Accelerator pedal position
S
5Engine temperature
Make the first two unit's functional be:
g
1(x)=c
1X, wherein c
1Be the constant that to determine.
Theoretical A indication operating range is D ≈ g
1(g
2(s
1)).
Make latter two unit functional be:
g
3(x，y)＝xy
Theoretical B indication operating range is D ≈ g
1(g
2(g
3(s
2, g
4(s
3)))).Each g that occurs
1All have different constants.
Can also be from sensor s
4The distance that estimation is travelled.For the illustration purpose, suppose in theoretical C, accelerator pedal position and acceleration linear dependence, and retarded velocity is only arranged owing to the wheel friction causes.
Make that next unit functional is the solution of following formula:
Then, the theoretical C indication distance of travelling is D ≈ g
1(
G2(g
5(s
4))).
In this way, three theories are divided into five unit functionals.
In this monitoring method and equipment, computerized algorithm can be tested the arrangement of these functionals and sensor, to find the most accurately combination.
The a little Advanced Edition of this method and apparatus of practical application can determine simultaneously that short-term test drives air resistance, engine friction force and the wheel friction force that vehicle is caused.
For example, the interpolation functional can be used for estimating the value of same vehicle different editions.For example, can use from passenger vehicle as a result the time, use the extrapolation functional to estimate the value of diesel truck.Need these functionals of predefined.
Described process has significant generality.Five same functionals are such as can being for the part from group's functional of the prediction aluminium corrosion such as sensor measurement corrosion current, relative humidity, surface moisture degree.
Example 2
This example relates to the corrosion condition of structure.
It should be noted that all five functionals that can analyze the manual car operating range with being used for and develop are used for diverse complication system, structure erosion, and need not to make modification.
Make g
1..., g
5As definition in the example 1.So:
g
1Purposes be the variation of unit, for example from the voltage volt of sensor measurement to relative humidity.Other purposes are to determine the corrosion condition (supposing in the plane) at angle and will totally corrode in addition related with the actual measured value of damage such as mass loss or cup depth.
g
2Purposes be to determine the damage of time integral from Damage.
The purposes of g3 is to obtain Damage from the function that wetness multiply by another variable.Another example is the effect that comprises the pH value.A lot of other purposes are arranged.
The purposes of g4 is the relation between wetted perimeter, surface area and the water droplet volume.A lot of other purposes are arranged.
The purposes of g5 is to carry out surface heating and cooling according to air heat and cooling.Another purposes is that the gasoloid that comprises electrolytic solution is penetrated in the cavity.
For the corrosion modeling to structure, can use case 1 in untapped four other unit functionals.
g
6(x，y)＝x+y
The heat that this functional is used for that the temperature difference is caused transmits modeling, average and interpolation among.There are other modes to do like this, for example utilize g (x, y)=cx+ (1-c) y and/or more senior interpolating method.
g
7(x)＝max(0，x-c
7)
Using this functional is in order to determine the wetness on surface according to local relative humidity.Another purposes is the delay modeling to using protective seam to cause.
Using this functional is for according to the moist time of relative humidity calculation.
Using this functional is the wetness that obtains crack or finedraw inside for the wetness from the outside.There are other modes to do like this, for example use
These nine unit functionals can be enough to the corrosion modeling of structure.
Three kinds of part of theories that they can be fit to corrode.Make s
1Be sensed relative humidity, s
2Be sensed aggressivity pollutant levels.The two all depends on the time.
Theoretical A: damage is proportional to the moist time.
D≈g
1(g
2(g
8(s
1)))
Theoretical B: damage is proportional to the time moist in the crack.
D≈g
1(g
2(g
8(g
9(g
7(s
1)))))
Theoretical C: damage is proportional to the deposition of aggressivity pollutant.
D≈g
1(g
2(g
3(g
8(s
1))，s
2))
Can be by similar fashion with the extra similar theory unit of resolving into functional.
Example 3
This example relates to the density of settling basin bottom mud.
This is another kind of diverse complication system.Can not make an amendment for six in nine functionals of example 2 and lend this example.These functionals are g
1And g
3To g
7For other functionals, get rid of g
8And g
9, because they are not had the essence needs, get rid of g
2, because the density of mud is instantaneous value but not integrated value, this distance from example 1 and 2 is different with damage.
Use at least one new unit functional, for the solution as the advection-diffusion equation of the given incoming fluid character of the function of time and mass rate.This can be expressed as:
Can develop the other unit functional and comprise other theories.
Example 4
This example has explained how the unit functional from for example example 2 can be combined into the functional group.
Each unit functional needs maximum constants.They can be made up in groups obtain correct final form.By grouping, can utilize multidimensional nonlinear least square curve fit (multidimensional nonlinear lest squares curve fitting) that they are fitted to sensor measurement.Can utilize the nonlinear optimization technology of standard, for example the method for Powell is done like this.
Can use different functionals by different modes.For example, can be with functional g
9Be specifically designed to extrapolation, generally be not used in nested combination, to allow the variation in environment/micro climate, g
6Can be used for the interpolation between general nested combination and diverse location and the environment.
Some functional combination may not be fit to.For example, consider the functional g of combination
i(g
j(... g
k(g
l(..., can impose restriction to limit repetition to k and l for every couple of k and l, avoid impossible figure and be essence reality.Such example can be:
Repeat: g
1(g
2=g
2(g
1Thereby, avoid g
2(g
1
Impossible figure: g
5(g
5Can be the perfect live part of the mathematical description of corrosion, but possibly can't be processed by numerical scheme, therefore avoid g
5(g
5
Essence reality: g
1(g
4(g
2(g
2Can be good fit to data, but may not correspond to any essence reality mechanism, therefore avoid g
2(g
2
Following table shows possible restriction set, and wherein " 1 " sign allows, and " 0 " expression is forbidden:
k＝1 k＝2 k＝3 k＝4 k＝5 k＝6 k＝7 k＝8
l＝1 0 0 0 0 0 1 0 0
l＝2 1 0 1 1 1 1 1 0
l＝3 1 1 1 1 1 1 1 1
l＝4 1 1 1 0 1 1 1 0
l＝5 1 1 1 1 0 1 1 1
l＝6 1 1 1 1 1 1 1 1
l＝7 1 1 1 1 1 1 0 0
l＝8 0 1 1 0 0 0 0 0
Should remove the repetition that allows in the table by from the nonlinear least square method match, checking the coupling posteriority ground (apostieri) of seeking minimum.
Possibility quantity is along with functional quantity increases soon, and nonlinear least square method may be slowly, the faster method that therefore can use elimination not make up.A kind of method is to use roughly monotonicity constraint.Suppose to have accepted ... g
i(g
j, found constraint c
iAnd c
j, and require test ... g
i(g
k(g
jThen utilize the previous c that finds
iAnd c
jAnd c
k=1 calculates ... g
i(g
k(g
jValue.If the result is dullness then refuse this layout roughly not, for example, if " the total rising "/" total decline " is positioned within the scope 0.1 to 10, then refuse this layout.If arrange and satisfy roughly monotonicity constraint, then carry out complete nonlinear least square method assessment.
The another kind of method of eliminating alternate item is the arrangement of applying unit functional rather than the combination of some or all unit functionals.This can realize very fast computerized algorithm, and along with the depth of nesting increases, it is faster rather than slower that this computerized algorithm becomes.If be modeled as the arrangement of theoretical (for example theoretical A, B and the C of example 1 and example 2) applying unit functional of all sons of functional group, this is a kind of effective mode.
Claims (32)
1. method of monitoring Evolution System, described method comprises the steps:
Obtain with from the relevant a plurality of sensor data streams of the output of the sensor of monitoring said system, the situation of at least one the described system of Sensor monitoring in the wherein said sensor, and the reason main body of the described situation of at least one Sensor monitoring in the wherein said sensor;
Make up iteratively a plurality of functional groups, each functional group is by the functional that is combined to form of concentrating the functional of selecting from fundamental functional;
By described sensor data stream being input in the described functional group, determine the output stream of each functional group;
From described a plurality of functional groups, select the functional group based on described output stream; And
Utilize selected functional group to come monitoring said system.
2. method according to claim 1 comprises the steps:
Obtain the status data stream relevant with the situation of described system; And
Selection has the functional group that flows the output stream of best-fit with described status data, as selected functional group.
3. method according to claim 1 and 2 is wherein from the combination of selected functional and the composite construction functional group of selected sensor data stream.
4. method according to claim 1 comprises the steps: to make up described fundamental functional collection by the analysis theory relevant with the behavior of described system.
5. method according to claim 1, the functional that wherein said fundamental functional is concentrated have dissimilar non-linear.
6. method according to claim 1, if wherein to a data flow operation, the functional of described fundamental functional collection has and is no more than a coefficient, if and/or wherein to two or more data flow operations, the functional of described fundamental functional collection does not have coefficient.
7. method according to claim 1 further comprises the steps:
Make up the optimum collection of functional group;
Monitor the new data value of described sensor data stream;
Utilize described new data value to optimize the coefficient that described optimum is concentrated the functional group, optimize the functional group to form; And
Concentrate one of described optimization functional group of selection as selected functional group from described optimum.
8. method according to claim 7 comprises the steps:
Monitor the renewal situation of described optimum collection; And
When described renewal situation occurring, by the described optimum collection of the incompatible reconstruct of optimal set of concentrating the search functional from described fundamental functional.
9. method according to claim 1 comprises the steps:
Provide the set of functional group from described functional collection;
To make up to provide other functional groups from functional group and other functionals of described functional group set;
Form the functional group set of upgrading from described functional group set and described other functional groups;
According to the best-fit situation the described functional group in the set of described renewal is sorted;
By abandoning the functional group that the poorest match is provided and the functional group that best-fit is provided by reservation, the quantity of the described functional group in the set of described renewal is reduced to below the maximum quantity;
Repeat described combination, formation, ordering and minimizing step, until arrive stop condition; And
From the functional group set of described renewal, select selected functional group.
10. method according to claim 1 wherein makes up described a plurality of functional group iteratively according to one or more rule, and described rule comprises one or more in following:
Restriction based on the repetition in the functional combination; Restriction based on the system action of functional group; Restriction based on the numerical value behavior of functional group; Functional is arranged but not the restriction of combination.
11. method according to claim 1 is wherein refused following functional group: described functional group provides the output stream of the expection feature that does not meet described Evolution System.
12. method according to claim 1 comprises the steps: to utilize selected functional group that the diagnostic message relevant with described system state is provided.
13. method according to claim 1 comprises the steps: to utilize selected functional group that the information of forecasting relevant with described system state is provided.
14. method according to claim 13, wherein by prediction in the future sensing data stream and the data stream of described prediction is input in the selected functional group described information of forecasting is provided.
15. method according to claim 14, wherein said future sensing data stream based on the data stream of previous sensing.
16. method according to claim 1, wherein said system is the first system, and wherein the functional group data of described the first system and the functional group data of second system are compared, and wherein, when the described functional group data of described the first and second systems are correlated with, when determining the selected functional group of described the first system, use the functional group data of described second system.
17. method according to claim 16, determine wherein whether described second system is in than described the first system and make progress faster state, if so, when determining the selected functional group of described the first system, use the functional group data of described second system.
18. method according to claim 1, wherein said system comprise a plurality of sensed and not sensed subsystems, and wherein by being used to monitor not sensed system from sensor data stream and the selected functional group of sensed subsystem.
19. method according to claim 18 is wherein revised selected functional group and/or sensor data stream by the extrapolation functional, to compensate the difference in the sensed and not sensed subsystem.
20. according to claim 18 or 19 described methods, wherein make up selected functional group and/or the sensor data stream of two or more sensed subsystems by the interpolation functional, so that selected functional group and/or the sensor data stream of not sensed subsystem to be provided.
21. according to claim 18 or 19 described methods, wherein obtain the selected functional group of not sensed subsystem from the first sensed subsystem, and wherein obtain the sensor data stream of described not sensed subsystem from the second sensed subsystem.
22. according to claim 18 or 19 described methods, comprise the steps: to determine the sensed subsystem with functional group data that is relative to each other, and utilize described correlation subsystem to determine similar subsystem in the described system.
23. method according to claim 1, wherein said system are the structural positions that experience is degenerated.
24. method according to claim 23, the reason main body of the degeneration by the sense position place and the described degeneration by the described position of sensing is wherein monitored the degeneration of described position.
25. according to claim 23 or 24 described methods, the position that wherein has similar functional group data is considered to degenerate in a similar manner.
26. according to claim 23 or 24 described methods, wherein be utilized as the selected functional group that the sense position that is regarded as and degenerates like the not sensed position class derives, monitor described structural described not sensed position.
27. method according to claim 1 wherein utilizes the discrete optimization algorithm that is applicable to uncertain polynomial time difficult problem to carry out the selection of functional group.
28. an equipment that is used for the monitoring Evolution System comprises:
The one group of sensor that is used for the parameter of monitoring said system, the situation of one of described sensor monitoring said system, and the reason main body of one of the described sensor described situation of monitoring; And
Processing module is used for:
From described one group of sensor receiving data stream;
Make up iteratively a plurality of functional groups, each functional group is the functional that is combined to form by the functional of selecting from the fundamental functional collection;
Apply described functional group to described data stream, to produce the output stream of each functional group;
Select the functional group based on described output stream; And
Utilize selected functional group monitoring said system.
29. equipment according to claim 28, wherein the form of sentencing sensor bunch of the ad-hoc location in described system provides described sensor together.
30. equipment according to claim 29 is wherein monitored a plurality of positions, the sensor that provides in this position bunch is provided in each position.
31. equipment according to claim 30, wherein the functional group data with described position compare mutually, in order to determine whether described system is similar in the behavior of described position.
32. according to claim 29,30 or 31 described equipment, wherein by using the position of not provide in the monitoring said system sensor from sensor data stream and the functional group data of sense position.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| AU2007902868 | 2007-05-29 | ||
| AU2007902868A AU2007902868A0 (en) | 2007-05-29 | Monitoring Methods and Apparatus | |
| PCT/AU2008/000754 WO2008144821A1 (en) | 2007-05-29 | 2008-05-29 | Monitoring methods and apparatus |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101790726A CN101790726A (en) | 2010-07-28 |
| CN101790726B true CN101790726B (en) | 2013-01-02 |
Family
ID=40074454
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008801009171A Active CN101790726B (en) | 2007-05-29 | 2008-05-29 | Monitoring methods and apparatus |
Country Status (8)
| Country | Link |
|---|---|
| US (1) | US8671065B2 (en) |
| EP (1) | EP2158549B1 (en) |
| JP (1) | JP5475649B2 (en) |
| CN (1) | CN101790726B (en) |
| AU (1) | AU2008255635B2 (en) |
| CA (1) | CA2689246C (en) |
| WO (1) | WO2008144821A1 (en) |
| ZA (1) | ZA200908376B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2012118390A1 (en) * | 2011-02-28 | 2012-09-07 | Critical Materials, Lda. | Structural health management system and method based on combined physical and simulated data |
| JP6227558B2 (en) | 2012-01-13 | 2017-11-15 | プロセス システムズ エンタープライズ リミテッド | Method, system, carrier media and product for monitoring a fluid treatment network |
| US9292023B2 (en) | 2012-09-12 | 2016-03-22 | International Business Machines Corporation | Decreasing the internal temperature of a computer in response to corrosion |
| EP3189436B1 (en) * | 2014-09-05 | 2018-07-18 | FTS Computertechnik GmbH | Computer system and method for security sensible applications |
| ES2600452B1 (en) * | 2016-03-17 | 2017-11-27 | Auscultia, S.L. | Method and system of predictive maintenance of buildings and structures. |
| US10671768B2 (en) * | 2016-07-01 | 2020-06-02 | The Boeing Company | Finite element modeling and analysis of crack propagation in multiple planes of a structure |
| WO2019027564A1 (en) * | 2017-08-03 | 2019-02-07 | Walmart Apollo, Llc | Automated item assortment system |
| JP7433171B2 (en) | 2020-09-08 | 2024-02-19 | 三菱重工業株式会社 | Display device, plant operation support system, and plant operation support method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6292810B1 (en) * | 1997-03-03 | 2001-09-18 | Richard Steele Richards | Polymorphic enhanced modeling |
| US6636842B1 (en) * | 2000-07-15 | 2003-10-21 | Intevep, S.A. | System and method for controlling an industrial process utilizing process trajectories |
| US6934931B2 (en) * | 2000-04-05 | 2005-08-23 | Pavilion Technologies, Inc. | System and method for enterprise modeling, optimization and control |
| US7216061B2 (en) * | 2004-08-25 | 2007-05-08 | Siemens Corporate Research, Inc. | Apparatus and methods for detecting system faults using hidden process drivers |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6453308B1 (en) | 1997-10-01 | 2002-09-17 | Aspen Technology, Inc. | Non-linear dynamic predictive device |
| US7725571B1 (en) * | 1999-05-24 | 2010-05-25 | Computer Associates Think, Inc. | Method and apparatus for service analysis in service level management (SLM) |
| US6865509B1 (en) * | 2000-03-10 | 2005-03-08 | Smiths Detection - Pasadena, Inc. | System for providing control to an industrial process using one or more multidimensional variables |
| AT412678B (en) * | 2002-09-30 | 2005-05-25 | Gerhard Dr Kranner | METHOD FOR COMPUTER-ASSISTED PREPARATION OF PROGNOSES FOR OPERATIONAL SYSTEMS AND SYSTEM FOR CREATING PROGNOSES FOR OPERATIONAL SYSTEMS |
| JP4348975B2 (en) * | 2003-03-20 | 2009-10-21 | トヨタ自動車株式会社 | Production line operating status analysis device, operating status analysis method, operating status analysis program, and operating status analysis system |
| US20050004833A1 (en) | 2003-07-03 | 2005-01-06 | Reaction Design, Llc | Method and system for integrated uncertainty analysis |
| WO2005013081A2 (en) * | 2003-08-01 | 2005-02-10 | Icosystem Corporation | Methods and systems for applying genetic operators to determine system conditions |
| JP4301049B2 (en) * | 2004-03-19 | 2009-07-22 | 富士電機ホールディングス株式会社 | Optimization method, optimization device, and optimization program |
| US7349823B2 (en) * | 2006-02-22 | 2008-03-25 | Sun Microsystems, Inc. | Using a genetic technique to optimize a regression model used for proactive fault monitoring |
-
2008
- 2008-05-29 AU AU2008255635A patent/AU2008255635B2/en active Active
- 2008-05-29 US US12/451,757 patent/US8671065B2/en active Active
- 2008-05-29 JP JP2010509624A patent/JP5475649B2/en active Active
- 2008-05-29 WO PCT/AU2008/000754 patent/WO2008144821A1/en active Application Filing
- 2008-05-29 EP EP08748015A patent/EP2158549B1/en active Active
- 2008-05-29 CN CN2008801009171A patent/CN101790726B/en active Active
- 2008-05-29 CA CA2689246A patent/CA2689246C/en active Active
-
2009
- 2009-11-26 ZA ZA200908376A patent/ZA200908376B/en unknown
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6292810B1 (en) * | 1997-03-03 | 2001-09-18 | Richard Steele Richards | Polymorphic enhanced modeling |
| US6934931B2 (en) * | 2000-04-05 | 2005-08-23 | Pavilion Technologies, Inc. | System and method for enterprise modeling, optimization and control |
| US6636842B1 (en) * | 2000-07-15 | 2003-10-21 | Intevep, S.A. | System and method for controlling an industrial process utilizing process trajectories |
| US7216061B2 (en) * | 2004-08-25 | 2007-05-08 | Siemens Corporate Research, Inc. | Apparatus and methods for detecting system faults using hidden process drivers |
Also Published As
| Publication number | Publication date |
|---|---|
| CA2689246A1 (en) | 2008-12-04 |
| AU2008255635B2 (en) | 2012-10-18 |
| EP2158549B1 (en) | 2012-12-05 |
| CN101790726A (en) | 2010-07-28 |
| EP2158549A4 (en) | 2011-10-19 |
| US20100169251A1 (en) | 2010-07-01 |
| AU2008255635A1 (en) | 2008-12-04 |
| EP2158549A1 (en) | 2010-03-03 |
| JP2010528379A (en) | 2010-08-19 |
| CA2689246C (en) | 2016-07-19 |
| JP5475649B2 (en) | 2014-04-16 |
| ZA200908376B (en) | 2010-08-25 |
| WO2008144821A1 (en) | 2008-12-04 |
| US8671065B2 (en) | 2014-03-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101790726B (en) | Monitoring methods and apparatus | |
| Nabian et al. | Efficient training of physics‐informed neural networks via importance sampling | |
| Zhang et al. | Probability and interval hybrid reliability analysis based on adaptive local approximation of projection outlines using support vector machine | |
| Sankararaman et al. | Bayesian methodology for diagnosis uncertainty quantification and health monitoring | |
| KR101105325B1 (en) | Method for Path-planning for Actual Robots | |
| Tran | Investigation of deterioration models for stormwater pipe systems | |
| CN111126611B (en) | High-speed traffic distribution simulation quantum computing method considering destination selection | |
| Aleisa et al. | For effective facilities planning: layout optimization then simulation, or vice versa? | |
| CN105117772B (en) | A kind of method for parameter estimation of multi-state System Reliability model | |
| Gordan et al. | Data mining based damage identification using imperialist competitive algorithm and artificial neural network | |
| CN105488317A (en) | Air quality prediction system and method | |
| Kim et al. | Navigation behavior selection using generalized stochastic Petri nets for a service robot | |
| Yoshida | Parameter study on optimal sampling planning based on value of information | |
| Sasmal et al. | Condition ranking and rating of bridges using fuzzy logic | |
| Haensel et al. | Collective risk minimization via a bayesian model for statistical software testing | |
| Kumar et al. | Stochastic adaptive sensor modeling and data fusion | |
| Sharifi et al. | A novel hybrid mechanistic-data-driven model identification framework using NSGA-II | |
| Sadrpour et al. | Real-time energy-efficient path planning for unmanned ground vehicles using mission prior knowledge | |
| Martin et al. | Just-in-time cooperative simultaneous localization and mapping | |
| Gopalakrishnan et al. | Use of neural networks enhanced differential evolution for backcalculating asphalt concrete viscoelastic properties from falling weight deflectometer time series data | |
| Lee | Digital Twins for Synchronisation of Production and Maintenance | |
| Shang | A Study of Deep Learning Neural Network Algorithms and Genetic Algorithms for FJSP | |
| Humeniuk et al. | Reinforcement learning informed evolutionary search for autonomous systems testing | |
| Hari Nugroho et al. | Risk‐based Inspection Scheduling Planning for Intelligent Agent in the Autonomous Fault Management | |
| Yu | Fault diagnosis and prognosis of hybrid systems using bond graph models and computational intelligence |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |